I’ll be the first to admit, as a researcher myself, I’m a fan of statistical methods for explaining and predicting the phenomenon we see in the real world.
But with the lousy polling in 2012, mostly showing a Mitt Romney edge at the end, coupled with the abysmal projections of a Hillary Clinton blowout in the Electoral College in 2016 — polling is now as useful as voodoo to predict politics.
Why is this?
In 2016, the pollsters measured all the wrong things — they totally ignored Millennials, totally ignored the working-class whites, and relied on ‘tried and true’ methods that wound up giving them the worst sort of information possible.
Moody’s Analytics, known for its perfect track record for calling presidential elections, has had its crystal ball shattered — even though they predicted that it could happen in 2015 with a Sanders or Trump candidacy, they still called the election overwhelmingly for Clinton.
All the major networks were in lockstep with a 300+ Electoral College win for Clinton.
In 2016, the pollsters measured all the wrong things -- they totally ignored Millennials, totally ignored the working-class whites, and relied on 'tried and true' methods that wound up giving them the worst sort of information possible.
So sure of the models, even Nate Silver’s warnings on Sunday that Clinton could be only one state away from an Electoral College loss, fell on deaf ears. His suggestion that Clinton might actually win the popular vote was almost scoffed by those tied to the models. Both predictions defied the polls, defied the odds, and quite frankly defied the ‘sensibilities’ of pollsters.
After all, how in the age old, tried-and-true models was a Democrat going to win a four-way race with both the popular and electoral vote?
Silver’s possibilities for a Democratic disaster turned into a reality when Clinton was denied from every single strategy of winning — the Florida ‘grand slam,’ the Great Lakes ‘firewall,’ and the Rust Belt ‘safety net.’ The only real positive Clinton maintained was the further ‘bluing’ of Nevada and holding swing states of Colorado and New Mexico.
All of which could have been accurately predicted by better polling models — they simply didn’t measure the blue-collar, white vote as a bloc that was shaping up to move in lock-step.
And after such a resounding defeat, who would have guessed that it would be Donald Trump that won without the popular vote — currently trailing almost a quarter of a million votes?
Silver’s last minute moment of pause still didn’t change his forecast that Clinton was going to win — and maybe it should have — he’d have been a hero of the statistical world today, not just part of the collective laughing stock.
But Larry Sabato is in particular going to want to hide under the proverbial rock for a few weeks — calling the presidential race (322-216, Clinton) and Senate-flip horridly (net D-gain of 4), and an outright disappointing call of the Democrats gaining +10-15 in the House (currently 239-192 — not even close).
In his mea culpa, Sabato writes:
“We heard for months from many of you, saying that we were underestimating the size of a potential hidden Trump vote and his ability to win. We didn’t believe it, and we were wrong … We have a lot to learn, and we must make sure the Crystal Ball never has another year like this. This team expects more of itself, and we apologize to our readers for our errors.” – Sabato Crystal Ball, Nov 9, 2016
In the end, every single polling model — from Fox News to CNN/ORC, Moody’s Analytics to Nate Silver’s FiveThirtyEight — all of these models are now fundamentally ‘broken.’
Is there really an answer when polling sciences, which have had fairly decent post-mortem accuracy over the past 30 years, are completely wrong?
Pollsters of the future, if anyone still even does them (Gallup got out of the presidential prediction game after their horrible showing in 2012), will have to focus on measuring the right demographics, measuring core issues better, and figuring out how to get ‘in touch’ with a far greater base of opinions.
Maybe we should mourn the death of election polling — though in reality we should expect more from our media and pollsters. Issues will always dominate fancy models predicting outcomes — what is America currently thinking?
Because in the end, it’s the issues that will be dealt with by our politicians — regardless of which one wins.
Photo Source: Charlie Neibergall / Associated Press